基于WorldView-2影像的土地利用信息提取方法对比及评价

基于2011年WorldView-2高分辨率遥感影像, 采取面向对象的分类方法和四种传统的基于像元的分类方法分别提取平潭县海坛岛中北部研究样区土地利用信息, 并以目视解译结果图为参考, 得到每种分类方法的总体分类精度, 且从数量分歧和分配分歧两方面对土地利用信息提取结果进行整体评价和单类别评价, 结果表明: (1)不同分类方法平均总体分类精度为75.00%, 其中最高的是面向对象法, 总体精度为84.25%, 分类总体精度最低的为最大似然法, 仅为62.00%. (2)面向对象分类方法具有最低的数量分歧, 为4.25%, 其次依次为神经网络法<支持向量机法<马氏距离法<最大似然法. 在分配分歧方面, 支持向量机方法其值最低, 为5.75%, 其次依次为最大似然法<神经网络法<马氏距离法<面向对象法. (3)在单类别精度评价中, 耕地的精度对影像整体分类结果影响最为显著, 其数量分歧比例大小依次为最大似然法(28.75%)>马氏距离法(21.50%)>支持向量机法(14.75%)>神经网络法(11.00%)>面向对象法(3.00%), 分配分歧比例大小依次为面向对象法(10.50%)>神经网络法(5.00%)>支持向量机法(1.50%)>最大似然法(0.50%)>马氏距离法(0.00%).

Comparison and Assessment of Land Use Information Extraction Methods Based on WorldView-2 Remote Sensing Image

Based on the WorldView-2 high resolution remote sensing image in 2011, this study uses object-based classification method and four traditional pixel-based classification methods to extract study area land use information respectively. Then, Visual interpretation map is functioned as reference map to acquire each classification methods overall accuracy and to assess the each classification result and each class type from the aspects of quantity disagreement and allocation disagreement. The result shows that: (1) The average overall classification accuracy is 75.00%. Among all the classification methods, the object-based classification method acquires the highest accuracy, 84.25%. The maximum likelihood classification method gets the lowest accuracy, 62.00%. (2) In all classification methods, the object-based classification method has obtained the lowest quantity disagreement, 4.25%. The others in sequence are as follows: neural net classification method < support vector machine method < mahalanobis distance method < maximum likelihood method. As to allocation disagreement, the support vector machine method has acquired the lowest value, 5.75%. The others in sequence are maximum likelihood method < neural net classification < mahalanobis distance method < object-based classification method. (3) As to separate class type, farmland does great influence on image''s overall classification accuracy, whose quantity disagreement sequence is the maximum likelihood method(28.75%) > mahalanobis distance method(21.50%) > support vector machine method(14.75%) > neural net method(11.00%) > object-based method(3.00%). As for allocation disagreement, the sequence is object-based method(10.50%) > neural net method(5.00%) > support vector machine method(1.50%) > maximum likelihood method(0.50%) > mahalanobis distance method(0.00%).